Section 4 AB-effects (metagenomes)

4.1 Load in variants

vcfToDataframe <- function(vcf.files, contig_mapping = contig_mapping, gff.df = gff.df) {
    require(vcfR)
    res <- list()
    for (file in vcf.files) {
        library(data.table)
        vcf.content <- vcfR::read.vcfR(file, verbose = FALSE)
        vcf.fix <- as.data.frame(vcf.content@fix)  # contains chr, position and substitution informations
        vcf.info <- vcfR::INFO2df(vcf.content)  # get INFO field, contains DP, AF informations
        if (nrow(vcf.fix) > 0) {
            # there are variants
            dat <- as.data.frame(cbind(vcf.fix[, c(1, 2, 4, 5, 6)], vcf.info[, c(1, 2)]))
            dat$majorAF <- sapply(dat$AF, minorAfToMajorAf)  # transforms e.g. AF of 0.1 to 0.9, 0.9 stays 0.9 and 0.5 stays 0.5
            dat$genome <- contig_mapping[match(dat$CHROM, contig_mapping$contig), ]$genome  # map chr information to genome name e.g. NHMU01000001.1 -> i48
            dat$genome_hr <- translateGenomeIdToFullName(tolower(dat$genome))
            dat$mouse.id <- substr(tools::file_path_sans_ext(basename(file)), 1, 4)
            dat$mouse.group <- translateMouseIdToTreatmentGroup(dat$mouse.id)
            dat$day <- as.integer(substr(basename(file), 6, 7))
            dat$phase <- binDaysByPhase(as.numeric(as.matrix(dat$day)))
            dat$phase_num <- binDaysByPhaseGroup(dat$day)
            dat$dp <- as.numeric(as.matrix(vcf.info$DP))
            # annotate overlay of gene
            dt.gff <- data.table(start = gff.df$start, end = gff.df$end, chr = as.character(as.matrix(gff.df$chr)), feature = gff.df$product)
            colnames(dat)[1:2] <- c("chr", "start")
            dat$start <- as.integer(as.matrix(dat$start))
            dat$chr <- as.character(as.matrix(dat$chr))
            dat$end <- dat$start
            dat2 <- as.data.table(dat)
            setkey(dt.gff, chr, start, end)
            annotated <- foverlaps(dat2, dt.gff, type = "within", mult = "first")
            res[[tools::file_path_sans_ext(basename(file))]] <- annotated  # add vcf df to list
        } else {
            message("Skipping")
        }
    }
    df <- as.data.frame(do.call(rbind, res))  # merge list to df
    return(df)
}

# load in reference information
gff.files <- Sys.glob("data/references/joined_reference_curated_ecoli/*.gff")
gff.df <- NULL
for (gff.file in gff.files) {
    message(gff.file)
    gff <- rtracklayer::readGFF(gff.file)
    # subset since different columns are present on gff files
    relevant <- data.frame(start = gff$start, end = gff$end, type = as.character(as.matrix(gff$type)), gene = as.character(as.matrix(gff$gene)), product = as.character(as.matrix(gff$product)), 
        chr = as.character(as.matrix(gff$seqid)))
    relevant$genome <- substr(basename(gff.file), 1, nchar(basename(gff.file)) - 4)
    gff.df <- rbind(gff.df, relevant)
}

## data/references/joined_reference_curated_ecoli/joined_reference_curated_ecoli.gff

# load in contig information
contig_mapping <- read.csv2("data/contig_mapping_new_ref.csv", sep = ";", header = T, stringsAsFactors = F)  # this file contains contig names of the 12 OligoMM genomes

# load in vcf files
vcf.files <- Sys.glob("out_philipp/all_vcf/*.vcf")
vcf.samples <- suppressWarnings(vcfToDataframe(vcf.files, contig_mapping, gff.df = gff.df))

## Skipping

vcf.samples$feature <- as.character(as.matrix(vcf.samples$feature))

vcf.samples[which(is.na(vcf.samples$feature)), ]$feature <- "outside ORFs"

vcf.samples$start <- NULL
vcf.samples$end <- NULL
vcf.samples$i.end <- NULL
colnames(vcf.samples)[3] <- "POS"

saveRDS(vcf.samples, file = "data/rds/omm_ab.rds")

4.2 AF frequency

p <- ggplot(vcf.samples, aes(AF, fill = genome)) + geom_histogram()
p <- p + facet_grid(mouse.id + mouse.group ~ genome + genome_hr)
p <- p + theme_classic() + xlab("AF") + ylab("occurence")
print(p)

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

Figure 4.1: AF of resequenced strains

p <- ggplot(vcf.samples, aes(majorAF, fill = genome)) + geom_histogram()
p <- p + facet_grid(mouse.id + mouse.group ~ genome + genome_hr)
p <- p + theme_classic() + xlab("AF") + ylab("occurence")
print(p)

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

Figure 4.2: major AF of resequenced strains

4.3 number of variants per samples

dat <- readRDS("data/rds/omm_ab.rds")
dat$dummy <- 1
dat.agg <- aggregate(dummy ~ mouse.id + mouse.group + day + phase, dat, sum)
DT::datatable(dat.agg)

4.3.1 number of variants per treatment group

p <- ggplot(dat.agg, aes(x = mouse.id, y = dummy, color = day))
p <- p + geom_jitter(shape = 4) + facet_grid(. ~ mouse.group, scales = "free_x")
p <- p + geom_boxplot() + theme_classic() + xlab("Mouse ID") + ylab("number of variants")
plotly::ggplotly(p)

Figure 4.3: number of variants of all 12 OMM genomes by mouse

4.4 Heatmap

All mice

dat <- readRDS("data/rds/omm_ab.rds")
dat$sample.id <- paste0(dat$mouse.id, "-", dat$day)
dat$variant.id <- paste0(dat$POS, "-", dat$REF, "-", dat$ALT)
data.wide <- dcast(dat, variant.id ~ sample.id, value.var = "AF")

## Warning in dcast(dat, variant.id ~ sample.id, value.var = "AF"): The dcast generic in data.table has been passed a data.frame and will attempt to redirect to the reshape2::dcast;
## please note that reshape2 is deprecated, and this redirection is now deprecated as well. Please do this redirection yourself like reshape2::dcast(dat). In the next version, this
## warning will become an error.

data.wide[is.na(data.wide)] <- 0
rownames(data.wide) <- data.wide$variant.id
data.wide$variant.id <- NULL

library(circlize)
library(ComplexHeatmap)

heat <- data.matrix(data.wide)
# limit to variants that are present in at least 10% of samples
heat_num <- rowSums(heat != 0)
heat2 <- heat[which(heat_num > ncol(heat)/10), ]

# limit to variants that have a high variance
heat_var_num <- matrixStats::rowVars(heat2)
heat3 <- heat2[which(heat_var_num > quantile(heat_var_num, 0.5)), ]

dat$dummy <- 1
annot.data <- aggregate(dummy ~ mouse.id + mouse.group + day + phase, dat, sum)
annot.data$sample.id <- paste0(annot.data$mouse.id, "-", annot.data$day)
heat3.mouse.id <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$mouse.id
heat3.day <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$day
heat3.mouse.group <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$mouse.group
heat3.phase <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$phase
heat3.phase2 <- ifelse(heat3.phase == "post-treatment", 6, NA)
ord = data.frame(day = heat3.day, mouse.id = heat3.mouse.id)

occ = as.data.frame(table(heat3.mouse.id))
ord$occ <- occ[match(ord$mouse.id, occ$heat3.mouse.id), ]$Freq
data.wide.sub <- dat[match(colnames(heat3), dat$sample.id), ]

col_fun = colorRamp2(c(0, 0.5, 1), c("white", "yellow", "red"))

qpcr <- read.table("qpcr.csv", header = T, sep = ";")
qpcr$universal <- NULL
rownames(qpcr) <- paste0(qpcr$mouse, "-", qpcr$day)
qpcr <- qpcr[, -c(1:5)]

qpcr <- apply(qpcr, 1, function(x) x/sum(x))

qpcr <- qpcr[, which(colnames(qpcr) %in% colnames(heat3))]
qpcr <- qpcr[, match(colnames(heat3), colnames(qpcr))]

# pdf('heat.pdf', width= 10, height = 10)
Heatmap(heat3, name = "AF", col = col_fun, border = TRUE, top_annotation = HeatmapAnnotation(num = anno_lines(colSums(heat3), smooth = TRUE, border = TRUE), ra = anno_barplot(t(qpcr), 
    bar_width = 1, gp = gpar(fill = 1:12), height = unit(3, "cm")), mouse = heat3.mouse.id, group = heat3.mouse.group, phase = heat3.phase, day = anno_simple(heat3.day, pch = heat3.phase2)), 
    cluster_columns = F, column_order = order(ord$occ, ord$mouse.id, ord$day), right_annotation = rowAnnotation(prev = anno_barplot(rowSums(heat3))), row_gap = unit(0, "mm"), column_gap = unit(0, 
        "mm"), column_split = heat3.mouse.group, column_names_gp = gpar(fontsize = 5), row_names_gp = gpar(fontsize = 3), show_row_dend = F, show_row_names = F, show_column_dend = F)

# dev.off()

All mice clustered

dat <- readRDS("data/rds/omm_ab.rds")
dat$sample.id <- paste0(dat$mouse.id, "-", dat$day)
dat$variant.id <- paste0(dat$POS, "-", dat$REF, "-", dat$ALT)
data.wide <- dcast(dat, variant.id ~ sample.id, value.var = "AF")

## Warning in dcast(dat, variant.id ~ sample.id, value.var = "AF"): The dcast generic in data.table has been passed a data.frame and will attempt to redirect to the reshape2::dcast;
## please note that reshape2 is deprecated, and this redirection is now deprecated as well. Please do this redirection yourself like reshape2::dcast(dat). In the next version, this
## warning will become an error.

data.wide[is.na(data.wide)] <- 0
rownames(data.wide) <- data.wide$variant.id
data.wide$variant.id <- NULL

library(circlize)
library(ComplexHeatmap)

heat <- data.matrix(data.wide)
# limit to variants that are present in at least 10% of samples
heat_num <- rowSums(heat != 0)
heat2 <- heat[which(heat_num > ncol(heat)/10), ]

# limit to variants that have a high variance
heat_var_num <- matrixStats::rowVars(heat2)
heat3 <- heat2[which(heat_var_num > quantile(heat_var_num, 0.5)), ]

dat$dummy <- 1
annot.data <- aggregate(dummy ~ mouse.id + mouse.group + day + phase, dat, sum)
annot.data$sample.id <- paste0(annot.data$mouse.id, "-", annot.data$day)
heat3.mouse.id <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$mouse.id
heat3.day <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$day
heat3.mouse.group <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$mouse.group
heat3.phase <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$phase
heat3.phase2 <- ifelse(heat3.phase == "post-treatment", 6, NA)
ord = data.frame(day = heat3.day, mouse.id = heat3.mouse.id)

occ = as.data.frame(table(heat3.mouse.id))
ord$occ <- occ[match(ord$mouse.id, occ$heat3.mouse.id), ]$Freq
data.wide.sub <- dat[match(colnames(heat3), dat$sample.id), ]

col_fun = colorRamp2(c(0, 0.5, 1), c("white", "yellow", "red"))

qpcr <- read.table("qpcr.csv", header = T, sep = ";")
qpcr$universal <- NULL
rownames(qpcr) <- paste0(qpcr$mouse, "-", qpcr$day)
qpcr <- qpcr[, -c(1:5)]

qpcr <- apply(qpcr, 1, function(x) x/sum(x))

qpcr <- qpcr[, which(colnames(qpcr) %in% colnames(heat3))]
qpcr <- qpcr[, match(colnames(heat3), colnames(qpcr))]

# pdf('heat.pdf', width= 10, height = 10)
Heatmap(heat3, name = "AF", col = col_fun, border = TRUE, top_annotation = HeatmapAnnotation(num = anno_lines(colSums(heat3), smooth = TRUE, border = TRUE), ra = anno_barplot(t(qpcr), 
    bar_width = 1, gp = gpar(fill = 1:12), height = unit(3, "cm")), mouse = heat3.mouse.id, group = heat3.mouse.group, phase = heat3.phase, day = anno_simple(heat3.day, pch = heat3.phase2)), 
    cluster_columns = T, right_annotation = rowAnnotation(prev = anno_barplot(rowSums(heat3))), row_gap = unit(0, "mm"), column_gap = unit(0, "mm"), column_names_gp = gpar(fontsize = 5), 
    row_names_gp = gpar(fontsize = 3), show_row_dend = F, show_row_names = F, show_column_dend = T)

# dev.off()

dat <- readRDS("data/rds/omm_ab.rds")
dat$sample.id <- paste0(dat$mouse.id, "-", dat$day)
dat$variant.id <- paste0(dat$POS, "-", dat$REF, "-", dat$ALT)
data.wide <- dcast(dat, variant.id ~ sample.id, value.var = "AF")

## Warning in dcast(dat, variant.id ~ sample.id, value.var = "AF"): The dcast generic in data.table has been passed a data.frame and will attempt to redirect to the reshape2::dcast;
## please note that reshape2 is deprecated, and this redirection is now deprecated as well. Please do this redirection yourself like reshape2::dcast(dat). In the next version, this
## warning will become an error.

data.wide[is.na(data.wide)] <- 0
rownames(data.wide) <- data.wide$variant.id
data.wide$variant.id <- NULL

library(circlize)
library(ComplexHeatmap)

heat <- data.matrix(data.wide)
# limit to variants that are present in at least 10% of samples
heat_num <- rowSums(heat != 0)
heat2 <- heat[which(heat_num > ncol(heat)/10), ]

# limit to variants that have a high variance
heat_var_num <- matrixStats::rowVars(heat2)
heat3 <- heat2[which(heat_var_num > quantile(heat_var_num, 0.5)), ]

dat$dummy <- 1
annot.data <- aggregate(dummy ~ mouse.id + mouse.group + day + phase, dat, sum)
annot.data$sample.id <- paste0(annot.data$mouse.id, "-", annot.data$day)
heat3.mouse.id <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$mouse.id
heat3.day <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$day
heat3.mouse.group <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$mouse.group
heat3.phase <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$phase
heat3.phase2 <- ifelse(heat3.phase == "post-treatment", 6, NA)
ord = data.frame(day = heat3.day, mouse.id = heat3.mouse.id)

occ = as.data.frame(table(heat3.mouse.id))
ord$occ <- occ[match(ord$mouse.id, occ$heat3.mouse.id), ]$Freq
data.wide.sub <- dat[match(colnames(heat3), dat$sample.id), ]

col_fun = colorRamp2(c(0, 0.5, 1), c("white", "yellow", "red"))

qpcr <- read.table("qpcr.csv", header = T, sep = ";")
qpcr$universal <- NULL
rownames(qpcr) <- paste0(qpcr$mouse, "-", qpcr$day)
qpcr <- qpcr[, -c(1:5)]

qpcr <- t(qpcr)

qpcr <- qpcr[, which(colnames(qpcr) %in% colnames(heat3))]
qpcr <- qpcr[, match(colnames(heat3), colnames(qpcr))]

# pdf('heat.pdf', width= 10, height = 10)
Heatmap(heat3, name = "AF", col = col_fun, border = TRUE, top_annotation = HeatmapAnnotation(num = anno_lines(colSums(heat3), smooth = TRUE, border = TRUE), ra = anno_barplot(t(qpcr), 
    bar_width = 1, gp = gpar(fill = 1:12), height = unit(3, "cm")), mouse = heat3.mouse.id, group = heat3.mouse.group, phase = heat3.phase, day = anno_simple(heat3.day, pch = heat3.phase2)), 
    cluster_columns = T, right_annotation = rowAnnotation(prev = anno_barplot(rowSums(heat3))), row_gap = unit(0, "mm"), column_gap = unit(0, "mm"), column_names_gp = gpar(fontsize = 5), 
    row_names_gp = gpar(fontsize = 3), show_row_dend = F, show_row_names = F, show_column_dend = T)

# dev.off()

4.5 Focus on mouse where we have many time points

dat <- readRDS("data/rds/omm_ab.rds")
dat$rep.group <- translateMouseIdToReplicateGroup(dat$mouse.id)
dat <- dat[which(dat$rep.group == "Full"), ]

dat$sample.id <- paste0(dat$mouse.id, "-", dat$day)
dat$variant.id <- paste0(dat$POS, "-", dat$REF, "-", dat$ALT)
data.wide <- dcast(dat, variant.id ~ sample.id, value.var = "AF")

## Warning in dcast(dat, variant.id ~ sample.id, value.var = "AF"): The dcast generic in data.table has been passed a data.frame and will attempt to redirect to the reshape2::dcast;
## please note that reshape2 is deprecated, and this redirection is now deprecated as well. Please do this redirection yourself like reshape2::dcast(dat). In the next version, this
## warning will become an error.

data.wide[is.na(data.wide)] <- 0
rownames(data.wide) <- data.wide$variant.id
data.wide$variant.id <- NULL

heat <- data.matrix(data.wide)
# limit to variants that are present in at least 10% of samples
heat_num <- rowSums(heat != 0)
heat2 <- heat[which(heat_num > ncol(heat)/10), ]

# limit to variants that have a high variance
heat_var_num <- matrixStats::rowVars(heat2)
heat3 <- heat2[which(heat_var_num > quantile(heat_var_num, 0.5)), ]

dat$dummy <- 1
annot.data <- aggregate(dummy ~ mouse.id + mouse.group + day + phase, dat, sum)
annot.data$sample.id <- paste0(annot.data$mouse.id, "-", annot.data$day)

heat3.mouse.id <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$mouse.id
heat3.day <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$day

ord = data.frame(day = heat3.day, mouse.id = heat3.mouse.id)

heat3.mouse.group <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$mouse.group
heat3.phase <- annot.data[match(colnames(heat3), annot.data$sample.id), ]$phase
heat3.phase2 <- ifelse(heat3.phase == "post-treatment", 6, NA)

col_fun = colorRamp2(c(0, 0.5, 1), c("white", "yellow", "red"))

# order the heatmap by treatment group

Heatmap(heat3, name = "AF", col = col_fun, border = TRUE, top_annotation = HeatmapAnnotation(num = anno_lines(colSums(heat3), smooth = TRUE, border = TRUE), day = anno_simple(heat3.day, 
    pch = heat3.phase2)), cluster_columns = F, column_order = order(ord$mouse.id, ord$day), column_split = heat3.mouse.group, column_names_gp = gpar(fontsize = 5), row_names_gp = gpar(fontsize = 8), 
    show_row_dend = F, show_row_names = F, show_column_dend = F)

4.6 Akkermansia Muciniphila

4.6.1 area plot 1

dat <- readRDS("data/rds/omm_ab.rds")
dat$variant.id <- paste0(dat$POS, "-", dat$REF, "-", dat$ALT)
dat <- dat[which(dat$chr == "Akkermansia_muciniphila"), ]
data.wide <- dcast(dat, day + mouse.id + mouse.group ~ variant.id, value.var = "AF")

## Warning in dcast(dat, day + mouse.id + mouse.group ~ variant.id, value.var = "AF"): The dcast generic in data.table has been passed a data.frame and will attempt to redirect to the
## reshape2::dcast; please note that reshape2 is deprecated, and this redirection is now deprecated as well. Please do this redirection yourself like reshape2::dcast(dat). In the next
## version, this warning will become an error.

data.wide[is.na(data.wide)] <- 0
colMax <- function(X) apply(X, 2, max)
dat_mat <- data.wide[, -c(1:3)]

# filter variants
data.wide.reduced <- cbind(data.wide[, c(1:3)], dat_mat[, which(colMax(dat_mat) > 0.5)])
# data.wide.reduced <- data.wide

dat2 <- melt(data.wide.reduced, id.vars = c("day", "mouse.id", "mouse.group"))

## Warning in melt(data.wide.reduced, id.vars = c("day", "mouse.id", "mouse.group")): The melt generic in data.table has been passed a data.frame and will attempt to redirect to the
## relevant reshape2 method; please note that reshape2 is deprecated, and this redirection is now deprecated as well. To continue using melt methods from reshape2 while both libraries
## are attached, e.g. melt.list, you can prepend the namespace like reshape2::melt(data.wide.reduced). In the next version, this warning will become an error.

dat3 <- dat2 %>% group_by(day, mouse.id) %>% mutate(Nor = value/sum(value))

set.seed(123)
col_list <- sort(unique(dat3$variable))
cols <- randomcoloR::randomColor(length(unique(dat3$variable)))

# Muller plot
p <- ggplot(dat3, aes(x = day, y = Nor, group = variable, fill = variable, label = ))
p <- p + geom_area(color = "black", size = 0.1)
p <- p + facet_wrap(~mouse.group + mouse.id, ncol = 3)
p <- p + theme_minimal() + theme(legend.position = "none")
p <- p + ylab("Fraction")
p <- p + scale_fill_manual(values = cols, breaks = col_list)
p <- p + geom_vline(xintercept = c(4, 18, 53, 67))
plotly::ggplotly(p)

4.6.2 line plot

dat <- readRDS("data/rds/omm_ab.rds")
dat$variant.id <- paste0(dat$POS, "-", dat$REF, "-", dat$ALT)
dat <- dat[which(dat$chr == "Akkermansia_muciniphila"), ]
data.wide <- dcast(dat, day + mouse.id + mouse.group ~ variant.id, value.var = "AF")

## Warning in dcast(dat, day + mouse.id + mouse.group ~ variant.id, value.var = "AF"): The dcast generic in data.table has been passed a data.frame and will attempt to redirect to the
## reshape2::dcast; please note that reshape2 is deprecated, and this redirection is now deprecated as well. Please do this redirection yourself like reshape2::dcast(dat). In the next
## version, this warning will become an error.

data.wide[is.na(data.wide)] <- 0
dat2 <- melt(data.wide, id.vars = c("day", "mouse.id", "mouse.group"))

## Warning in melt(data.wide, id.vars = c("day", "mouse.id", "mouse.group")): The melt generic in data.table has been passed a data.frame and will attempt to redirect to the relevant
## reshape2 method; please note that reshape2 is deprecated, and this redirection is now deprecated as well. To continue using melt methods from reshape2 while both libraries are
## attached, e.g. melt.list, you can prepend the namespace like reshape2::melt(data.wide). In the next version, this warning will become an error.

set.seed(123)
col_list <- sort(unique(dat3$variable))
cols <- randomcoloR::randomColor(length(unique(dat3$variable)))
p <- ggplot(dat2, aes(x = day, y = value))
p <- p + geom_line(aes(group = variable), alpha = 0.2)
p <- p + theme_minimal()
p <- p + facet_wrap(~mouse.group + mouse.id, ncol = 3)
p <- p + geom_vline(xintercept = c(4, 18, 53, 67))
plotly::ggplotly(p)